The present invention relates to a nonvolatile memory device, a method of manufacturing the device, and a method of driving the device, and more particularly, to a nonvolatile memory in which the coupling ratio of the memory cells is increased without increasing cell size, through the structure and operation of a "program assist plate," thereby lowering the operating voltage and increasing the operating speed of the device. The invention may be used in many different types of nonvolatile memory devices, including NAND, NOR, AND, DINOR and other devices.
In a NOR-type electrically erasable programmable read-only memory(EEPROM), two facing memory cells share one bitline contact and one source line, and the memory cells in a row are connected to one bitline. Thus, it is difficult to highly integrate the NOR-type structure, although its high cell current allows it to operate at high speeds.
In a NAND-type structure, two cell strings share one bitline contact and one source line. In one cell string, a plurality of cell transistors are connected in series to the bitline. Accordingly, the NAND-type structure can easily obtain a high level of integration, but it is typically slower than the NOR-type structure due to its low cell current. Because the NAND-type memory cell can be more highly integrated than the NOR-type memory cell, it is generally preferable to employ the NAND-type memory structure for increasing the capacity of a memory device. However, this invention is not limited to application in only NAND type devices
The EEPROM NAND string structure and the basic operation of the NAND-type EEPROM are described below, referring to the accompanying drawings.
FIG. 1 is a plan view showing the layout with respect to one string in a typical NAND-type nonvolatile memory device, and FIG. 2 is an equivalent circuit diagram of the structure shown in FIG. 1.
Referring to FIGS. 1 and 2, each string of a NAND-type nonvolatile memory device is formed by sequentially connecting a string selection transistor S1, a plurality of cell transistors C1, . . . , Cn and a source selection transistor S2 in series between the bitline B/L and a source line S/L in an area represented by a width x and a length y.
FIG. 3A is a plan view of a transistor cell used in forming each string of the nonvolatile memory device, and FIG. 3B is a sectional view taken along line I-I' of FIG. 3A.
In FIG. 3A, reference numeral 26 indicates a mask pattern for forming an active region, reference numeral 24 indicates a mask pattern for forming a control gate, and reference numeral 22 indicates a mask pattern for forming a floating gate.
Referring to FIG. 3B, each transistor cell C1, . . . , Cn of FIG. 1 in the string consists of a floating gate 32, a control gate 34 and a N-type source/drain 36, which are sequentially deposited on a P-type semiconductor substrate 30, with an interdielectric layer inserted therebetween. The programming, erasing and reading of a NAND-type nonvolatile memory device having this structure is described below.
The NAND-type nonvolatile memory is programmed by tunneling an electric charge from a channel region of the cell transistor to the floating gate thereof, to thereby store information. For example, if information is to be programmed or stored in the first transistor cell C1, power supply voltage Vcc is applied to the gate of string select transistor S1, thereby turning on string select transistor S1, and 0 V is applied to the gate of source select transistor S2, thereby turning off source select transistor. With reference to FIG. 3B, a programming voltage Vpgm is applied to the control gate if 34 of the first transistor cell C1, to thereby generate tunneling. Accordingly, an electric charge in the channel region of the substrate 30 moves to the floating gate 32, to thereby change the threshold voltage Vth of the first transistor cell C1.
After programming, transistor cell C1 will have (approximately) one of two different threshold voltages depending on the charge transferred to the floating gate 32. The first and second threshold voltages may correspond to either a "1" or "0" in a two-state memory device. In a multi-state memory device more than two threshold voltages may be used, thereby storing more than one bit per cell.
A read operation is used to determine the programmed state of the NAND memory cell. For example, referring to FIG. 2, when reading information stored in the first cell transistor C1, the bitline B/L is precharged with a predetermined voltage between approximately 1 V.about.Vcc. Then, Vcc is applied to each control gate of the string select transistor S1, the source select transistor S2 and unselected cell transistors C2, . . . , Cn, (i.e., each cell transistor except for C1) to thereby turn-on the transistors. Approximately 0 V is applied to the control gate of the selected first cell transistor C1, which is between a first threshold voltage of approximately -3 V when a "1" is stored in the cell and a second threshold voltage of approximately 1 V when a "0" is stored in cell C1. Thus, if the first cell transistor C1 is turned on, and a current is sensed between the bitline B/L and source line S/L, the state of the first cell transistor C1 is determined as "1". However, if the first cell transistor C1 is turned off, and no (or very little) current is sensed between the bitline B/L and the source line S/L, the state of the first cell transistor C1 is determined as "0". Alternatively, no current could correspond to a "1" and a sensed current could correspond to a "0."
The erasing operation is performed by tunneling an electric charge from the floating gate 32 to the channel region of the substrate 30 (FIG. 3B), thereby erasing information stored in the cell. For example, referring to FIG. 2, when information is to be erased from the first cell transistor C1, the cell string is placed in a floating state by disconnecting it from the bitline B/L and the source line S/L by turning off the string select transistor S1 and source select transistor S2. A voltage of 0 V is applied to all wordlines of a selected block of memory cells C1, C2, . . . Cn. Further, referring to FIG. 3B, an erase voltage Verase is applied to the substrate 30, thereby generating tunneling from the floating gate 32 to the substrate 30. Thus, the electric charge on the floating gate 32 is moved to the substrate 30, thereby changing the threshold voltage of the selected memory cells.
In the operation of the nonvolatile memory device described above, a high-voltage of approximately 20 V is required to program or erase the memory cells by Fowler-Nordheim (referred to as "F-N") tunneling. A charge pumping circuit is required to supply a high voltage for programming and erasing, which results in increased chip size and power consumption. Accordingly, in order to increase the density of a nonvolatile memory device, it is important to increase the efficiency of both erasing and programming, and thereby lower the power requirements for Vpgm and Verase.
In order to enhance the operating characteristics without lowering the reliability of the nonvolatile memory device, the capacitance of the structure corresponding to the interdielectric layer deposited between the control gate 34 and the floating gate 32 must be increased, and the program/erase voltage must be lowered. The capacitance may be increased by reducing the thickness of the interdielectric layer or increasing the contact area of the control gate 34 and the floating gate 32. If the capacitance is increased by reducing the thickness of the interdielectric layer, the data retention capability of the nonvolatile memory device is reduced, and the insulation of the interdielectric layer may be broken during programming and erasing. In addition, the process for producing an interdielectric layer of reduced thickness is difficult. However, a method has recently been developed for increasing the contact area between the control gate 34 and the floating gate 32.
FIG. 4 is a plan view showing a layout of a conventional NAND-type nonvolatile memory device, disclosed in IEDM Tech. Dig. 1994, pp. 61-64, which is incorporated by reference herein. This article discloses a structure and method for obtaining high-integration and increased capacitance with respect to the interdielectric layer by increasing the effective surface area.
In FIG. 4, reference numeral 40 denotes a mask pattern for defining an active region, reference numeral 42 denotes a mask pattern for forming a floating gate, reference numeral 44 denotes a mask pattern for forming a control gate, and reference numeral 46 denotes a mask pattern for forming a bitline contact. The mask pattern 42 for forming the floating gate completely overlaps with the mask pattern 40 for defining an active region. That is, the floating gate is self-aligned on the active region, which leads to high-integration.
FIG. 5 is a sectional view taken along line II-II' of FIG.4, where reference numeral 50 denotes a semiconductor substrate, reference numeral 52 denotes a floating gate, reference numeral 54 denotes an interdielectric layer, reference numeral 56 denotes a control gate, and reference numeral 58 denotes an isolation film. In this nonvolatile memory device, the area of the interdielectric layer 54 between the floating gate 52 and the control gate 56 is determined only by the width of the active region of the cell transistor. The active regions are the portions of the substrate 50 located between adjacent isolation films 58. Note that the floating gate 52 is not formed on the isolation film 58. Accordingly, the capacitance related to the interdielectric layer 54 is lowered, and the resulting device requires a high voltage for programming and erasing.
To solve the above problems, the thickness of the floating gate 52 of FIG. 5 is increased. However, the thick floating gate 52 structure causes two problems. First, when the control gate 56, the interdielectric layer 54, and the floating gate 52 are patterned in accordance with this process, a vertically high interdielectric layer 54 must be formed on the sidewalls of the thick floating gate 52. In addition, it is difficult to etch the thick floating gate 52 in the source/drain region (not shown) of a cell transistor.